What is the difference between 'experimental' and 'beta' data sources?

What’s the difference between these in terms of:

  1. Stability: can both be used in production use cases?
  2. Completeness: do these sources offer enough features to be useful in practice?
  3. Support: what kind of support does Palantir provide? Does Palantir accept feature requests for these? Does Palantir provide (critical) incident support in case of regressions?